improved evaluation of cover crop species by growth and

HAL Id: hal-00886527https://hal.archives-ouvertes.fr/hal-00886527

Submitted on 1 Jan 2010

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Improved evaluation of cover crop species by growth androot factors

G. Bodner, M. Himmelbauer, W. Loiskandl, H.-P. Kaul

To cite this version:G. Bodner, M. Himmelbauer, W. Loiskandl, H.-P. Kaul. Improved evaluation of cover crop speciesby growth and root factors. Agronomy for Sustainable Development, Springer Verlag/EDP Sci-ences/INRA, 2010, 30 (2), �10.1051/agro/2009029�. �hal-00886527�

https://hal.archives-ouvertes.fr/hal-00886527

https://hal.archives-ouvertes.fr

Agron. Sustain. Dev. 30 (2010) 455–464c© INRA, EDP Sciences, 2009DOI: 10.1051/agro/2009029

Research article

Available online at:www.agronomy-journal.org

for Sustainable Development

Improved evaluation of cover crop species by growth and root factors

G. Bodner1*, M. Himmelbauer2, W. Loiskandl2, H.-P. Kaul1

1 Institute of Agronomy and Plant Breeding, Department of Applied Plant Sciences and Plant Biotechnology, University of Natural Resources and Applied LifeSciences Vienna, Gregor Mendel Straße 33, 1190 Vienna, Austria

2 Institute of Hydraulics and Rural Water Management, Department of Water, Atmosphere and Environment, University of Natural Resources and Applied LifeSciences Vienna, Muthgasse 18, 1190 Vienna, Austria

(Accepted 20 July 2009)

Abstract – Cover crops are plants that are integrated in the crop rotation between two cash crops. The main objectives of cover cropping areorganic matter input, mitigation of nitrate leaching and reduction of soil erosion. These benefits will only be achieved efficiently if the selectedcover crop species are adapted to local environmental conditions and appropriate for the defined agro-ecological target. Therefore, a mainlimitation in cover cropping is the lack of a comprehensive species description. An improved cover crop characterization could be achievedwith quantitative parameters derived from growth functions. Here, we show the use of this approach to assess plant traits relevant for erosioncontrol by cover cropping. An experiment with four cover crop species (phacelia, vetch, rye and mustard) was performed over two years at asemi-arid site in Eastern Austria. Canopy cover was measured four times over the vegetation period. Root length density measurements weremade to 40 cm soil depth before winter. Canopy dynamics were characterized by parameters from the asymptotic Gompertz function and froman extended logistic model that includes a parameter for decay after maximum coverage. Our results show that vetch had the lowest early vigorafter dry conditions at sowing, with +45% longer time to attain maximum growth rate (parameter tmax) than the other species. Drought duringthe later autumn growing period led to the highest reduction in maximum canopy cover (parameter ymax) for phacelia (–41%). The rootingpattern was assessed by parameters from the exponential distribution function of Gerwitz and Page. The most intense rooting near the soilsurface (parameter L0) was found for phacelia (9.7 cm cm−3). Vetch had the lowest L0 (4.6 cm cm−3) but highest root allocation to deeper soillayers. Mustard combined high average values in ymax (76%) and L0 (6.3 cm cm−3) with a stable growth over both years. The potential strengthsof phacelia and vetch were more dependent on the particular year. Rye showed a stably high L0 (8.6 cm cm−3), but had only a low average valueof ymax (55.1%). The quantitative parameter sets we derived for plant traits required for erosion control improved cover crop comparison andanalysis of their local adaptation. Based on this extended species description our approach allows a better evaluation of cover crops and can beused for the optimization of management and decision support.

cover crops / growth functions / species description / rooting pattern / rainfall distribution

1. INTRODUCTION

Cover cropping is a widely used agro-environmental prac-tice to reduce negative effects of post-harvest fallowing duringautumn and winter. Cover crops reduce nitrate leaching (e.g.Thorup-Kristensen, 2001; Vidal and Lopez, 2005, Rinnofneret al., 2008) and improve soil physical properties (MacRae andMehuys, 1985). Both living crops and mulch cover protect thesoil surface from the impact of raindrops and can reduce runoffand soil erosion by more than 95% compared with fallow (e.g.Meyer et al., 1999; Zuazo and Pleguezuelo, 2008). Quintonet al. (1997) studied the impact of increasing canopy coverageon soil loss. They found that the greatest reduction occurredfor canopy covers greater than 30% with a maximum reduc-tion reached when 70% of the soil was covered by vegetation.

* Corresponding author: [email protected]

Besides soil cover, the cover crop root system is of particu-lar importance to improve soil quality parameters such as bio-logical activity (Schutter and Dick, 2002), aggregate stability(Liu and Bomke 2005) and hydraulic properties (Carof et al.,2007; Bodner et al., 2008). Furthermore, a high and stableaboveground growth needs sufficient water and nutrient sup-ply via the root system.

While previous research mainly focused on effects of covercrops in relation to environmental or agronomic parameters,there is a lack of studies that provide a detailed characteriza-tion of the available cover crop species themselves. However,the quantification of cover crop growth dynamics is essentialfor the optimization of their management to obtain the desiredagro-environmental effects.

Different mathematical functions with biologically mean-ingful parameters have been used to describe growth processes

Article published by EDP Sciences

http://dx.doi.org/10.1051/agro/2009029

http://www.agronomy-journal.org

http://www.edpsciences.org

456 G. Bodner et al.

Figure 1. Temperature and precipitation during the field experiment compared with the long-term (30 years) averages, with high (2004) vs. low(2005) rainfall at cover crop seeding in August and regular (2004) vs. irregular (2005) rainfall during the autumn growing period.

(e.g. Werker and Jaggard, 1997) and root distribution (e.g.Feddes and Raats, 2004). Based on the model parameters, dif-ferences between plant species can be characterized quanti-tatively and environmental influences on growth performancecan be studied. This study presents an approach for improvedcharacterization of cover crop species based on parametersfrom plant growth and root distribution functions. The ap-proach is used to assess canopy coverage and rooting traitsof different cover crop species, which are two important prop-erties of cover crop plants in relation to erosion control. Weanalyze the sensitivity of the species in terms of these traitsfor two years of contrasting rainfall pattern under semi-aridconditions. This exemplifies the use of quantitative approachesin species description, to obtain more targeted recommenda-tions of appropriate cover crop species and ensure their agro-environmental benefits.

2. MATERIAL AND METHODS

2.1. Study site and experimental set-up

The data used for the present study are from a field experi-ment in the pannonic region of Eastern Austria (48◦12’N and16◦34’E). Climatically the site is characterized by semi-aridconditions with an average annual precipitation of 491 mm, amean annual temperature of 9.1 ◦C, an average wind speed of3 m s−1 and an average global radiation of 17.1 MJ m−2. Theexperiment is located on a sloping chernozem soil with a siltcontent between 45.5 and 50.4%. Due to these characteristicsthe field is susceptible to soil erosion and therefore represen-tative of sites where cover cropping is typically used to avoidsoil degradation.

Weather data were recorded by an automated ADCONweather station at the experimental field. Two consecutiveyears were evaluated. They showed distinct differences inweather conditions. Such differences are typically found dur-ing the cover crop vegetation period in the pannonic climate.Figure 1 shows mean precipitation and temperature compared

with the long-term averages for the cover crop growing periodin both years.

The field experiment consisted of four cover crops fromdifferent plant families and with different root characteris-tics. Phacelia (Phacelia tanacetifolia Benth. cv. Vetzrouska)is a non-winter-hardy cover crop from the Hydrophyllaceaefamily with a taproot system concentrated in the upper soil(Hampl, 1996). Hairy vetch (Vicia villosa L. cv. Beta) is awinter-hardy legume species with a primary root branchinginto several lateral roots of similar diameter (Kutschera et al.,2009). Rye (Secale cereale L. cv. Picasso) is also a winter-hardy cover crop with the typical dense adventitious root sys-tem of grasses (Kutschera et al., 2009). Mustard (Sinapis albaL. cv. Caralla) is a non-winter-hardy Brassicacea species witha strong taproot (Hampl, 1996). Seeding rates were 10 kg ha−1

for phacelia, 90 kg ha−1 for vetch, 120 kg ha−1 for rye and10 kg ha−1 for mustard. In accordance with the Austrian agro-environmental program ÖPUL, cover crops were sown on20 August. Plots (60 m2) were arranged in a randomized com-plete block design with three replications. Cover crops fol-lowed barley in both years. Management of the main crop wasthe same in all plots. The main crop was fertilized with 50 kgN ha−1 while the cover crops did not receive any fertilization.For the present study, the focus was on the main growing pe-riod of the cover crops from seeding until December, when thenon-winter-hardy species were killed by frost.

2.2. Canopy cover measurements

Canopy cover of the crops was measured four times duringthe growing period by image analysis of digital pictures ac-cording to Karcher and Richardson (2005). Three digital pho-tos were taken per plot from a constant height of one meterabove the ground. Image analysis for percent ground coverwas performed based on color discrimination using the soft-ware SigmaScan Pro5. Additionally, aboveground dry matterof the cover crops was measured on the last canopy cover sam-pling date from a sample area of 1 m2.

Improved evaluation of cover crop species by growth and root factors 457

2.3. Root sampling and analysis

Root samples were taken using the soil core method (Böhm,1979) in mid-December at the end of the cover crop vegeta-tion period when the plants were likely to have reached theirmaximum growth before winter. The auger used for samplinghad an inner diameter of 7 cm. Due to limited auger length,sampling depth was restricted to the upper 40 cm. However,this can be expected to cover the soil layers where most rootsare concentrated and therefore allow a sufficient description ofcover crop rooting traits with relevance for soil structure. Twosoil cores were taken per plot, one directly below the plant andone between two rows. The samples (1540 cm3) were then di-vided into three sub-samples from 0–10 cm, 10–20 cm and20–40 cm soil depth. The procedure of root length measure-ment is described in detail by Himmelbauer et al. (2004). Inshort, roots were separated from soil by a hydro-pneumaticelutriator and debris as well as dead roots were removed fromthe samples. Root length was then measured by image analy-sis of stained roots using WinRhizo 4.1 (Régent Instruments,Quebec).

2.4. Growth models for canopy cover

Two model approaches were used to derive parameterscharacterizing the growth pattern of cover crop canopies. Theclassical Gompertz function assumes an asymptotic growth toa maximum coverage and is given by (e.g. Pegelow et al.,1977):

yi = ymax,G exp{− exp

(kG

(ti − tmax,G

))}(1)

where yi (%) is ground cover at day ti (d) after sowing,ymax,G (%) is the maximum coverage, kG is the growth rate(d−1) and tmax,G (d) is the time until growth rate is maximum.

Frequently, a decrease in canopy cover from a maximumvalue can be observed due to leaf wilting or senescence.Werker and Jaggard (1997) presented a modification of theGompertz model by including a decay term. The model isgiven by:

yi = ymax,WJ exp

(µmin(ti − tmax,WJ) − µmin

kWJ(1 − e−kWJ(ti−tmax,WJ))

)

(2)where ymax,WJ (%) again is the maximum coverage, µmin (d−1)is the decay rate, tmax,WJ (d) in this case is the time until groundcover reaches its maximum and kWJ (d−1) is the rate that de-termines how fast the initial growth approaches µmin.

2.5. Root distribution function

Using measured root length density data, root distributionwas characterized by the exponential function of Gerwitz andPage (1974), which describes the decrease in rooting densitywith depth by:

RLDi = L0e−azi (3)

where RLDi (cm cm−3) is root length density at soil depthzi (cm), L0 (cm cm−3) is root length density at the soil sur-face (z = 0) and a (dimensionless) is a parameter describingthe decrease in root length density with depth.

2.6. Data analysis, calculation and evaluationof distinctive plant parameters

The evaluation of measurements followed the stepwise ap-proach used by Schabenberger and Pierce (2002) for the anal-ysis of factorial experiments with non-linear response. In afirst step, the whole data set was submitted to a mixed modelanalysis of variance using PROC MIXED in the SAS softwarepackage. The correlation structure among repeated measure-ments in the statistical analysis was described by a first-orderautoregressive model (Piepho et al., 2004). The objective ofthe initial analysis of variance was to derive the factors of ma-jor influence (i.e. year, species, block and measurement datein the case of canopy cover). This identifies those subsets ofthe whole data set that differ significantly from each other andthus require a separate growth model for their proper charac-terization.

The subsequent description of the data subsets was doneby curve fitting of growth (Eqs. (1) and (2)) and root distribu-tion functions (Eq. (3)) using non-linear regression by the SASprocedure PROC NLIN.

In the case of canopy cover, the two different growth func-tions (Eqs. (1) and (2)) were first evaluated concerning theirgeneral appropriateness to describe the respective data sets.When both functions provided a significant fit with a uniqueparameter set, their appropriateness was decided using theAkaike Information Criterion (AIC) as a goodness of fit pa-rameter.

The subsequent analysis of the parameters from the fittedgrowth functions was performed with respect to (i) their dis-tinctiveness between the four cover crop species and (ii) theirstability over the two years for each crop species. As therewas no common growth function shared by all data sets (cf.Sect. 3.2) and thus no possibility of a single ANOVA-basedanalysis incorporating both effects (species and year), two dif-ferent approaches were used to answer the respective ques-tions.

The distinctiveness of model parameters between the covercrop species was determined by common ANOVA and leastsignificant differences (LSD). The analysis was done sepa-rately for each year. This is justified because (i) the initialanalysis of variance indicated significant year differences, andconsequently (ii) only within one year did cover crop speciesshare a common growth model.

The second question, i.e. growth performance stability ofeach cover crop species over the two years, required the def-inition of a common growth model shared by the respectivespecies in both years. The decision on the adequate model foreach species was taken based on the Akaike Information Crite-rion (AIC). For the subsequent evaluation of the intra-speciesparameter sensitivity to the year, a sum of squares reductiontest (SSR-test) was chosen. This approach is suitable as the


species-wise comparison of growth performance is only tar-geting the year effect and does not imply any interaction effect(Schabenberger and Pierce, 2002). The SSR test statisticallycompares a “full model” where all parameters are year-specificwith the same model sharing one or several common param-eter values over both years. Shared parameter values expressa stable growth performance in the respective trait. The com-parison is done stepwise starting with a completely reducedmodel (i.e. no difference in all growth parameters between theyears) and successively allowing for an increasing number ofyear-specific parameters. The process is stopped at the num-ber of fixed parameters where no significant difference occursbetween the full and reduced model at P < 0.05. This finalmodel reveals sensitive and non-sensitive growth parametersof the respective cover crop species. The F-value for the sta-tistical test is given by:

F =RSSr−RSSf

dff−dfr

RSSfdfres,f

P = Pr (Fdff−dfr,dfres,f � F) (4)

where RSSr is the residual sum of squares of the reducedmodel, RSSf is the residual sum of squares of the full model,dff and dfr are the model degrees of freedom of the full andreduced model, respectively, and dfres,f are the error degreesof freedom of the full model. The resulting F-value is com-pared with the tabulated F-value at (dff-dfr) numerator degreesof freedom and dfres,f denominator degrees of freedom, givingthe probability level P.

The rooting system was analyzed using the root distribu-tion function parameters of equation (3). Again, evaluation ofspecies’ distinctiveness in root parameters was performed bycommon ANOVA and least significant differences, while yearstability/plasticity in the rooting pattern of the individual covercrops was assessed using a SSR test.

3. RESULTS AND DISCUSSION

A field experiment was performed in order to develop animproved quantitative method for the evaluation of differ-ent cover crop species. The performance of four cover crops(phacelia, vetch, rye and mustard) was studied over two years,in terms of their canopy coverage and rooting density, two im-portant plant traits for erosion control and soil structure stabi-lization. Using measured data for these two plant variables,quantitative parameters were derived from growth and rootdistribution functions to assess the vigor and the stability ofthe investigated species under the local semi-arid conditions.

3.1. Climatic characteristics of the growing periods

The two experimental years showed substantial differencesin their climatic growing conditions during the cover crop veg-etation period, particularly in relation to rainfall distribution(Fig. 1). In 2004, dry conditions occurred at the time of seed-ing and germination of the cover crops. August precipitation

was 66% lower compared with the long-term average. Dur-ing the later cover crop growing period from mid-Septemberto mid-November there was regular rainfall. In 2005, by con-trast, August was characterized by very high precipitation,accounting for 67% of the total rainfall during the covercrop vegetation period before winter. After the 29 Septem-ber until 5 December, only 11 mm of rain fell, resulting inseverely dry conditions during the main cover crop growingperiod. In 2004 the climatic water balance deficit (cumulativerainfall minus cumulative potential evapotranspiration) was−132.4 mm. This high deficit resulted from the dry conditionsuntil mid-September when the evaporative demand of the at-mosphere was still high. In 2005 the cumulative deficit waslower (–19.2 mm). However, between October and Decemberthe low rainfall resulted in a continuously negative balance be-tween precipitation and evapotranspiration.

3.2. Growth function performance

The analysis of variance of canopy cover data indicated asignificant interaction between year × sampling date × covercrop (P = 0.02). This defines the data subsets for non-linear fitting to consist of year-wise cover crop-specific growthcurves (Fig. 2).

It should be noted that high standard errors in 2004 reflecta pronounced inhomogeneity of the stands, which developedfrom suboptimal dry soil conditions at seedbed preparation. Atthe toe-slope of the experimental field, with higher clay con-tent, this resulted in clods, and consequently a poor seed-soilcontact. The subsequent delay in emergence and early growthat the toe-slope was not regained during the entire growingperiod.

The three-parameter Gompertz model gave a satisfactorydescription of canopy growth for all cover crops in 2004. In2005 the Werker & Jaggard function was required to capturethe observed final decay. For vetch, with a delayed onset ofgrowth and a sustained increase in canopy cover until the finalobservation date before winter, the Werker & Jaggard modeldid not converge in 2004. In 2005 the Gompertz model did notfit for rye due to the large final decrease, as well as for mustarddue to early reduction in soil cover from the maximum thatwas reached already in October. For those species where bothgrowth models showed a significant fit (i.e. 2004: phacelia,rye and mustard; 2005: phacelia and vetch), the more suitablemodel was decided based on the value of the Akaike Informa-tion Criterion (AIC) as a statistical parameter for fitting qual-ity. The AIC indicated that the Gompertz logistic curve fittedbetter for all cover crops in 2004, while the growing pattern of2005 was better described by the growth and decay function ofWerker & Jaggard. The two years differed substantially in thetime until crop emergence. For the ANOVA evaluation of yeareffects on canopy cover at a given date, measurements weretaken at similar times after sowing. From the point of viewof growth model fitting, however, differentiated measurementdates would have been preferable, mainly in the early growingperiod.


Figure 2. Growth functions fitted to the measured ground cover (means ± standard error) of different cover crop species and in two years ofcontrasting rainfall pattern. The asymptotic Gompertz model describes the observed dynamics accurately, except for those cases where droughtinduced a substantial final reduction of canopy cover in 2005. These cases required an additional decrease parameter in the function (Werker &Jaggard model).

Several empirical functions are available to characterize thelogistic growth dynamics that are frequently observed in na-ture (e.g. Tsoularis and Wallace, 2002). Werker and Jaggard(1997) extended some of these functions to include a decreas-ing branch after the maximum in order to analyze phenomenasuch as the effects of drought, disease or herbicide damageon the plant canopy. These extended models therefore include

an additional parameter for decay. The higher complexity ofthe model may increase stability problems of the iterative pa-rameter estimation procedure, mainly related to the selectionof adequate initial values (Robert et al., 1999; Schabenbergerand Pierce, 2002; Yin et al., 2003). The growth functions usedin our study (Gompertz model and extended Gompertz modelby Werker & Jaggard) were fitted with a Marquardt algorithm


Table I. Estimated parameter values (with standard errors in paren-theses) of the Gompertz model describing the growth of canopy cov-erage of different cover crops in 2004. For significant effects in theANOVA at P < 0.05, comparison of means is indicated by lower-case letters. Those species sharing the same letters are not signifi-cantly different.

ymax,G (%) tmax,G (d) kG (d−1)Phacelia 79.9 (7.0) 36.4 (3.8)a 0.098 (0.048)Vetch 64.1 (8.9) 47.9 (5.2)b 0.091 (0.024)Rye 60.6 (3.1) 31.3 (4.4)a 0.073 (0.023)Mustard 72.1 (6.4) 32.5 (3.7)a 0.161 (0.081)

P = 0.17 P = 0.03 P = 0.61

Legend: ymax,G = maximum canopy coverage, kG = growth rate, tmax,G =

time until maximum growth rate (Eq. (1)).

in PROC NLIN of SAS. Although only a limited number ofmeasurement points were available from the field experiment,the iteration method converged for all cases. The uniquenessof the parameter estimates was further tested by using differ-ent initial values. The same parameter values were obtained.This indicated that the iteration algorithm indeed converged ata global minimum (Schabenberger and Pierce, 2002).

From the point of view of cover crop species characteri-zation for soil protection, parameters that describe the earlyvigor in surface coverage and the total canopy growth poten-tial in autumn are most relevant. The additional information ofthe Werker & Jaggard model shows differences in the sensitiv-ity to adverse conditions, such as drought, which lead to somecanopy decrease in the late season. A slight canopy reductionin the late season, however, is less important in the context oferosion control by cover cropping. A sufficiently high surfacecover is generally maintained by the living canopy and mulchof falling leaves.

A more detailed analysis of cover crop interactions withthe environment would suggest the use of mechanistic growthmodels. However, such models require a large number of mea-surement data for an accurate parameterization of the mod-eled plant-soil-atmosphere system and validation of simula-tion results. This need for multiple parameters is a constraintof mechanistic models compared to a growth function analy-sis. In comparative species description, it must be possible toevaluate a large number of species and sites (multi-locationtrials). This imposes a limit on measurement frequency anddetail. Therefore, we consider growth functions that can beeasily parameterized with sufficient accuracy as most adequatefor this purpose.

3.3. Comparison of canopy coverage dynamics betweendifferent cover crops

The model parameters of the fitted functions were sub-sequently analyzed for inter-specific differences between thecover crops. Tables I and II show the results of the analysis ofvariance for differences between the species in the Gompertzparameters for 2004 and the Werker & Jaggard’s parametersfor 2005, respectively.

Table II. Estimated parameter values (with standard errors in paren-theses) of the Werker & Jaggard’s model parameters describing thegrowth of canopy coverage of different cover crops in 2005. For sig-nificant effects in the ANOVA at P < 0.05, comparison of means isindicated by lower-case letters. Those species sharing the same lettersare not significantly different.

ymax,WJ (%) tmax,WJ (d) µmin (d−1) kWJ (d−1)Phacelia 47.0 (5.3)a 66.1 (6.2) –0.197 (0.042)a 0.006 (0.003)Vetch 93.5 (2.1)b 62.8 (3.8) –0.011 (0.001)c 0.052 (0.014)Rye 49.6 (4.6)a 54.0 (5.7) –0.061 (0.014)ab 0.023 (0.008)Mustard 76.8 (1.2)b 45.9 (1.8) –0.034 (0.003)bc 0.131 (0.017)

p < 0.01 P = 0.14 P = 0.02 P = 0.08

Legend: ymax,WJ = maximum canopy coverage, tmax,WJ = time until max-imum canopy coverage, µmin = decay rate, kWJ = growth rate (Eq. (2)).

Foley (1999) suggested rapid emergence and vigorous earlygrowth under a wide range of environmental conditions as animportant breeding objective for cover crops. This is partic-ularly important at sites with potential soil moisture shortageafter cash crop harvest. This was the case in 2004 in our study,when susceptible species could be substantially impaired intheir early stages. In this year cover crops differed significantlyin tmax,G, the time to reach the maximum growth rate, kG. Vetchhad a tmax,G that was 15 days longer compared with the averageduration for the other species (33 d). Clark (2007) pointed outthat dry conditions could reduce germination and retard earlygrowth of vetch. This can be related to a higher seed weightof vetch compared with the other species investigated and thusmore water required to initiate germination.

Besides the characteristic delay of vetch to reach its maxi-mum growth rate, there was no distinct inter-species variabilityin canopy cover dynamics under conditions of regular rainfallduring the autumn growing period. All crops achieved a finalcanopy cover higher than 60%, with phacelia having the high-est maximum coverage (79.9%). Thus, our results suggest thatin spite of a potential delay in early growth, as in the caseof vetch, this can be recovered without detriment for the fi-nal soil cover, provided that there is still sufficient growingtime before winter (Teasdale et al., 2004) and favorable growthconditions in terms of rainfall. For aboveground biomass therewas a wider range of species differentiation, although statisti-cally not significant, with phacelia (1577 kg ha−1) and mustard(1450 kg ha−1) at the higher end, and rye (907 kg ha−1) andvetch (859 kg ha−1) at a low level of dry matter production.This shows that, in contrast to its soil coverage, vetch obvi-ously did not recover its initially delayed growth in terms ofbiomass accumulation.

Furthermore, a delayed early canopy coverage can be a sub-stantial disadvantage concerning avoidance of runoff and ero-sion irrespective of the height of final ground cover. Assuminga lower threshold of 30% ground cover (Quinton et al., 1997),mustard achieved this value 28 days after sowing, phacelia 35,rye 37 and vetch only 52 days after sowing, respectively.

In the case of sufficient soil moisture for fast germina-tion, as occurred in 2005, crops showed a homogeneous


establishment. In this year water-limiting conditions devel-oped during a prolonged dry period in autumn. This led toa differentiation in maximum cover (ymax,WJ) as well as a fi-nal decrease in canopy coverage (µmin) between the cover cropspecies (Tab. II).

Phacelia and rye were those species less tolerant to the 2005rainfall pattern with dry conditions during the main growingperiod. Both achieved a low ymax,WJ below 50%. The rapidcanopy decrease (µmin) shows premature leaf senescence andabscission due to the low autumn rainfalls.

Parameters describing the increasing branch of the growthcurve to the year-specific maximum value, i.e. kWJ and tmax,WJ,did not differ significantly between the species in 2005. How-ever, the lower ymax,WJ of phacelia and rye led to a longer timeto reach the threshold of 30% soil surface cover, which wasreached 19 days after sowing by vetch, 22 by mustard, 30 byrye and 39 by phacelia, respectively.

For phacelia, to our knowledge there are no detailed studiesconcerning its susceptibility to drought. The observed canopyreduction under dry conditions was probably related to a lim-ited plasticity in root penetration to depth (cf. Sect. 3.5). Ryeis generally considered a cover crop that provides high soilcover and tolerance to dry conditions (e.g. Ingels et al., 1998;Sattell et al., 1998). This contradicts the low canopy cover andbiomass observed in both years in our study. The erectophileleaf orientation of the monocotyledonous rye compared tomore planophile dicotyledonous leaves may result in a lowercanopy cover. Furthermore, we observed brown rust infectionof the rye plants in both years. This might have negatively af-fected both photosynthetic capacity as well as water-use effi-ciency (Paul and Ayres, 1984). The low canopy cover in 2005also corresponded to a very low biomass of rye (712 kg ha−1)compared with the other cover crops. Phacelia was at an inter-mediate biomass level with 1135 kg ha−1, similar to mustard(1415 kg ha−1), in spite of the lower maximum canopy cover

Vetch and mustard both performed well under the dry au-tumn conditions in 2005 with a high canopy cover. Their tol-erance to water shortage is also reflected in a lower µmin com-pared with phacelia and rye. Vetch also showed a dry matteraccumulation significantly superior to all other species in thisyear (2338 kg ha−1). These findings are in agreement with ob-servations reported by Clark (2007) on hairy vetch, and by Ganet al. (2007) on mustard, who mentioned a comparatively hightolerance to water stress for these two species.

3.4. Comparison of year sensitivity in canopy coverageof individual cover crops

Cover cropping also requires that species are sufficientlystable in the desired growth traits under situations of differentrainfall distribution and water availability. This was assessedby the sum of squares reduction test (cf. Eq. (4)). This testreveals which stable growth function parameters were sharedover both years, and those parameters with high year variabil-ity under the different weather conditions. Vetch shared theGompertz model in both years and was compared using the

Table III. Results of the sum of squares reduction (SSR) testfor canopy cover models. Parameters/parameter sets indicate thosemodel parameters which need to be given a year-specific value inorder to achieve a non-significant difference to the full model casewhere all parameters have year-specific values.

Parameters / F Pparameter sets*

Phacelia ymax,WJ 3.42 0.07Vetch tmax,G 4.68 0.06

Ryeµmin+kWJ 1.90 0.40µmin+tmax,WJ 1.04 0.31kWJ+tmax,WJ 1.75 0.23

Mustard tmax,WJ 3.08 0.09

Gompertz parameters. The other cover crops were evaluatedusing the Werker & Jaggard model parameters.

Table III shows the F- and p-values of the sum of squaresreduction test. At the given p-values the reduced model doesnot differ significantly from the full model. At this stage, theparameters/parameter sets indicated in Table 3 have a year-specific value, while all other parameters of the respectivegrowth functions (Eqs. (1) and (2)) share a common value inboth years. Growth parameters having a common value ex-press stability in the respective trait in spite of the differentrainfall availability of the two years.

In case of phacelia, a year specific Ymax was required. Themaximum canopy cover of phacelia was highly dependent onyearly growth conditions, being reduced from 79.9% in 2004to only 47.0% in 2005 by a lack of rainfall during the period ofmajor biomass growth. The dynamics of growth, as expressedby the growth rate and the time to reach the year-specific max-imum cover, were less variable in response to the climatic con-ditions.

Vetch and mustard, on the other hand, were less variableconcerning their maximum ground cover before the end ofthe growing season. The differing weather conditions affectedmore the temporal dynamics expressed in the tmax parametersof the two models. For vetch tmax,G changed from 48 days in2004 to 22 days in 2005. This indicates the early growth sen-sitivity due to delayed germination and reduced early growthvigor under the dry sowing conditions of 2004. For mustard itshows some year-specific effect on the time to attain maximumsoil cover (tmax,WJ), being 74 days in 2004 and only 46 days in2005, respectively. Plants often shorten their vegetative growthperiod under water stress (e.g. Bernier and Périlleux, 2005),which was apparently the case for mustard in 2005, alreadyreaching its maximum at the beginning of October. However,from the point of view of soil protection, annual variabilityis more critical if it occurs during the early stages, as in thecase of vetch, when canopy cover is still low (Quinton et al.,1997). Mustard therefore can be regarded as a reliable covercrop providing a stable high cover of more than 70%. Its highgrowth rate guarantees a fast initial coverage of the soil sur-face. Haramoto and Gallandt (2004), reviewing cover crop usefor weed control, also pointed out that the fast soil coverage isa general strength of Brassica cover crops.


Figure 3. Root length density (RLD) means, standard errors and fitted Gerwitz & Page root distribution functions. Highest surface near RLD(parameter L0) was achieved by rye (2004) and phacelia (2005). Roots were more concentrated in the upper layer in 2004, while in 2005decrease in RLD with depth (parameter a) was less.

Rye showed a higher variability in its canopy growth dy-namics in the two years than the other cover crops. The lowestdifference from the full model was obtained with year-specificvalues for kWJ and µmin. Only the maximum canopy coveragewas stable over both years, although on quite a low level. Thishigh variability in growth dynamics and a generally low max-imum coverage would thus discourage the use of a rye mono-culture for soil surface protection.

Under the semi-arid conditions at the site biomass growthof the cover crops was generally lower compared with valuesindicated in the literature for Central Europe (Lütke Entrup,1986). We consider water to be the main limiting factor at thesite. The distinct effects on the investigated cover crop speciesdepending on the timing of water stress were shown clearly.Some additional influence of different intra-specific reactionsof the species to the given nutrient status of the soil cannot beexcluded. However, it should be of minor importance becauseof the generally high nutrient status of the chernozem soil andthe nutrient input via the conventionally fertilized cash crops.

3.5. Rooting pattern of cover crops

Cover crop rooting traits contribute essentially to their agro-environmental functions (e.g. Thorup-Kristensen, 2001). Theroot system also ensures sufficient water and nutrient supplyfor a stable cover crop growth. We described the cover cropspecies by root distribution function parameters of the Gerwitz& Page model (Eq. (3)). They provide two advantages for our

Table IV. Estimated parameter values (with standard errors in paren-theses) of the Gerwitz & Page model for root length density distribu-tion of cover crops. For significant effects in the ANOVA at P < 0.05,comparison of means is indicated by lower-case letters. Those speciessharing the same letters are not significantly different.

2004 2005

L0 (cm cm−3) a (–) L0 (cm cm−3) a (–)Phacelia 7.2 (1.8)ab –0.075 (0.021) 12.2 (1.2)a –0.040 (0.008)Vetch 4.9 (1.6)b –0.128 (0.055) 4.6 (1.0)b –0.022 (0.013)Rye 10.6 (2.2)a –0.087 (0.026) 6.6 (1.0)ab –0.029 (0.011)Mustard 5.2 (2.3)b –0.090 (0.035) 7.4 (1.1)ab –0.034 (0.012)

purpose. Its first model parameter (L0) captures the intensity ofrooting near the soil surface. This allows a good characteriza-tion of the cover crops’ potential contribution to soil structurestabilization in the uppermost soil layer. The second parame-ter (a) describes the decrease in rooting density with depth andpoints to the plants’ access to water and nutrients from deepersoil layers. It could therefore provide a background for the in-terpretation of species’ aboveground performance and stabil-ity. Measured root length density and distribution functions areshown in Figure 3; estimates of the corresponding Gerwitz &Page parameters are given in Table IV.

Analysis of variance of the Gerwitz & Page model parame-ters revealed major species differentiation in near-surface rootlength density (L0). In both years vetch had the lowest valuefor L0. The low L0 of vetch is in agreement with observations


Table V. Results of the sum of squares reduction (SSR) test for theroot model. Parameters/parameter sets indicate those model parame-ters which need to be given a year-specific value in order to achieve anon-significant difference to the full model case where all parametershave year-specific values.

Parameters / F Pparameter sets

PhaceliaL0 0.45 0.52a 1.41 0.26

Vetch a 0.01 0.97Rye – 2.20 0.16

MustardL0 1.28 0.28a 0.36 0.56

of Kutschera et al. (2009) for several legume root systemswhich show comparatively large root diameters, but less in-tense rooting of the soil. The cover crop species with highestL0 were rye in 2004 and phacelia in 2005. Mustard differedsignificantly to rye only in 2004. In 2005 mustard also hadan intense near-surface rooting, but without significant differ-ences from either phacelia or rye. Liu et al. (2005) demon-strated that the intense rooting of non-leguminous cover cropshad a particularly large effect on the improvement of aggre-gate stability. In their review, Zuazo and Pleguezuelo (2008)referred to intense topsoil rooting as a main contribution toerosion control, which confirms the findings of Sarrantonioand Gallandt (2003) for rye.

The decay parameter a of the Gerwitz & Page modelshowed a large small-scale variability and did not differ sig-nificantly between species. It mainly reflected the general in-fluence of the year, revealing a typical plant root response todrought: under conditions of regular rainfall during the veg-etation period (2004) roots concentrate near the surface andsharply decrease with depth (amean = 0.10). By contrast, irreg-ular rainfall with prolonged dry periods (2005) fosters deeproot growth to improve water uptake (e.g. Kage et al., 2004).This is expressed in a lower average value of a (0.03). Rootingdensities tend from an exponential to a more linear decreasewith depth.

Table V shows the results for root parameter sensitivity tothe year for the individual cover crop species revealed by thesum of squares reduction (SSR) test.

Rye was insensitive to the year, i.e. even the totally reducedmodel with both parameters having common values in bothyears did not differ significantly from the full model where allparameters are year-specific. This stable intense rooting of ryeemphasizes that the main agro-environmental contribution ofthis cover crop is related to its belowground traits, particularlyconsidering the inferior aboveground growth that we found inour study.

Vetch required the integration of a year-specific value forthe parameter a. This indicates a high plasticity in root pro-liferation to depth in response to soil moisture for this covercrop. The root systems of phacelia and mustard were satisfac-torily described over the two years if a year-specific value isattributed to at least one parameter. For phacelia the respec-

tive P-values suggested a higher year sensitivity in L0, and formustard in the parameter a.

Root plasticity, as suggested for vetch and to a lower extentalso for mustard, improves plant response to water stress (Belland Sultan, 1999). This is reflected by the lower susceptibilityof these crops to canopy reduction after the dry autumn periodin 2005. The proliferation of their root system to deeper layersallowed a higher water uptake from the soil.

Those cover crop species with the highest density in near-surface rooting (rye, phacelia) can be expected to have sub-stantial benefits for soil stabilization and amelioration (ag-gregate stability, soil microbiological activity), although thesecrops are more susceptible to drought in their abovegroundtraits, as expressed in their high µmin in 2005.

4. CONCLUSION

Cover cropping is a widely used agro-environmental man-agement practice in Europe. A detailed species description ofcover crops, however, is still lacking. Our study demonstratesthe use of quantitative parameters from growth and root distri-bution functions for an improved cover crop characterization.This approach was applied to the canopy cover and rootingpattern of four cover crop species (phacelia, vetch, rye andmustard) under semi-arid conditions. The obtained parameterswere used to characterize the species in terms of plant traitsrelevant for erosion control. Two main distinctions in canopycover among the species could be revealed from the functionparameters: delay in early vigor after dry sowing conditions(captured by the parameter tmax) and limitation of maximumcoverage due to low autumn rainfall (expressed in the parame-ter ymax). These main distinctive parameters could be obtainedfrom the asymptotic Gompertz function. Using an extendedfunction also enabled the assessment of the decrease in canopycover in late growth stages. Vetch was most affected by drysowing conditions in its early vigor. This could be shown by a+45% increase in the time to attain maximum growth rate (pa-rameter tmax) compared with the other species. Phacelia wasmost susceptible to the lack of autumn rainfall, reducing max-imum coverage (parameter ymax) by –41 %. Root characteriza-tion using an exponential distribution function could capturespecies differences in near-surface root density (L0) and de-crease in rooting (a) with depth. Phacelia had the highest av-erage L0 with 9.7 cm cm−3. Vetch had the lowest value for L0(4.6 cm cm−3) but showed a high capacity of deep root alloca-tion under dry autumn conditions (low decrease in parametera). Mustard combined high parameter values for ymax (mean:76%) and L0 (mean: 6.3 cm cm−3), a fast initial coverage, andstable growth performance over both years. Rye also showedstability in ymax and L0. High values, however, were only ob-tained in L0 (mean 8.6 cm cm−3), while ymax was low in bothyears (mean 55.1%).

The results demonstrate that our growth function-basedapproach could provide an improved cover crop species de-scription. The function parameters allowed a quantitative com-parison among the species and an assessment of their growthstability under different environmental conditions.


The species description proposed in this study provides thebasis for further developments towards a multi-environment,multi-trait database of cover crops. For this purpose additionaldata from a wider range of soil and climatic conditions shouldbe included in this approach and further parameters related toother agro-environmental targets should be defined, e.g. for or-ganic matter input, nitrate leaching prevention or weed control.The availability of comprehensive quantitative parameter setsof cover crop characteristics will be an important contributionto management optimization and improved decision support.

REFERENCES

Bell D.L., Sultan S.E. (1999) Dynamic phenotypic plasticity for rootgrowth in Polygonum: A comparative study, Am. J. Bot. 86, 807–819.

Bernier G., Périlleux C. (2005) A physiological overview of the geneticsof flowering time control, Plant Biotechnol. J. 3, 3–16.

Bodner G., Loiskandl W., Buchan G., Kaul H.-P. (2008) Natural andmanagement-induced dynamics of hydraulic conductivity along acover-cropped field slope, Geoderma 146, 317–325.

Böhm W. (1979). Methods of studying root systems, Ecological Studies33, Springer-Verlag, Berlin Heidelberg.

Carof M., Tourdonnet S., Coquet Y., Hallaire V., Roger-Estrade J. (2007)Hydraulic conductivity and porosity under conventional and no-tillage and the effect of three species of cover crop in northernFrance, Soil Use Manage. 23, 230–237.

Clark A. (2007) Managing cover crops profitably, Sustainable agriculturenetwork handbook series, bk. 9, 144.

Feddes R.A., Raats P.A.C. (2004) Parameterizing the soil-water-plantroot system, in: Feddes R.A., de Rooij G.H., van Dam J.C. (Eds.),Unsaturated-zone modelling, Progress, challenges and applications,Kluwer Academic Publishers, pp. 95–144.

Foley M. (1999) Genetic approach to the development of cover crops forweed management, J. Crop. Prod. 2, 77–93.

Gan Y., Malhi S.S., Brandt S., Katepa-Mupondwa F., Kutcher H.R.(2007) Brassica juncea canola in the northern great plains: re-sponses to diverse environments and nitrogen fertilization, Agron.J. 99, 1208–1218.

Gerwitz A., Page E.R. (1974) An empirical mathematical model to de-scribe plant root systems, J. Appl. Ecol. 11, 773–781.

Hampl U. (1996) Gründüngung, Leopold Stocker Verlag, Graz-Stuttgart.

Haramoto E.R., Gallandt E.R. (2004) Brassica cover cropping for weedmanagement: A review, Renew. Agr. Food Syst. 19, 187–198.

Himmelbauer M.L., Loiskandl W., Kastanek F. (2004) Estimating length,average diameter and surface area of roots using two different imageanalysis systems, Plant Soil 260, 111–120.

Ingels C.A., Bugg R.L., Thomas F.L. (1998) Cover crop species and de-scriptions, Chapter 2, in: Ingels C.A., Bugg R.L., McGourty G.T.,Christensen L.P. (Eds.), Cover Cropping in Vineyards: A Grower’sHandbook, University of California, Division of Agriculture andNatural Resources, Publication 3338, pp. 8–39.

Kage H., Kocher M., Stützel H. (2004) Root growth and dry matter par-titioning of cauliflower under drought stress conditions: measure-ment and simulation, Eur. J. Agron. 20, 379–394.

Karcher D.E., Richardson M.D. (2005) Batch analysis of digital imagesto evaluate turfgrass characteristics, Crop Sci. 45, 1536–1539.

Kutschera L., Lichtenegger E., Sobotik M. (2009) Wurzelatlasder Kulturpflanzen gemäßigter Gebiete mit Arten desFeldgemüsebaues, DLG-Verlags-GmbH, Frankfurt am Main.

Liu A., Ma B.L., Bomke A.A. (2005) Effects of cover crops on soil aggre-gate stability, total organic carbon, and polysaccharides, Soil Sci.Soc. Am. J. 69, 2041–2048.

Lütke Entrup E. (1986) Feldfutter und Zwischenfruchtbau, in:Oehmichen J. Pflanzenproduktion, Band 2: Produktionstechnik,Paul Parey, pp. 508–565.

MacRae R.J., Mehuys G.R. (1985) Effect of green manuring on the phys-ical properties of temperate-area soils, Adv. Soil Sci. 3, 71–94.

Meyer L.D., Dabney S.M., Murphree C.E., Harmon W.C., GrissingerE.H. (1999) Crop production systems to control erosion and reducerunoff from upland silty soils, T. ASAE 42, 1645–1652.

Paul N.D., Ayres P.G. (1984) Effects of rust and post-infection droughton photosynthesis, growth and water relations in groundsel, PlantPathol. 33, 561–569.

Pegelow E.J., Taylor B.B., Horrocks R.D., Buxton D.R., Marx D.B.,Wanjura D.F. (1977) The Gompertz function as a model for cottonhypocotyl elongation, Agron. J. 69, 875–878.

Piepho H.P., Büchse A., Richter C. (2004) A mixed modelling approachfor randomized experiments with repeated measures, J. Agron.Crop Sci. 190, 230–247.

Quinton J.N., Edwards G.M., Morgan R.P.C. (1997) The influence of veg-etation species and plant properties on runoff and soil erosion: re-sults from a rainfall simulation study in south east Spain, Soil UseManage. 13, 143–148.

Rinnofner T., Friedel J.K., de Kruijff R., Pietsch G., Freyer B. (2008)Effect of catch crops on N dynamics and following crops in organicfarming, Agron. Sustain. Dev. 28, 551–558.

Robert N., Huet S., Hennequet C., Bouvier A. (1999) Methodology forchoosing a model for wheat kernel growth, Agronomie 19, 405–417.

Roush W.B., Branton S.L. (2005) A comparison of fitting growth modelswith a genetic algorithm and nonlinear regression, Poultry Sci. 84,494–502.

Sarrantonio M., Gallandt E. (2003) The role of cover crops in NorthAmerican cropping systems, J. Crop Prod. 8, 53–74.

Sattell R., Dick R., Karow R., Kaufman D., Hemphill D., Luna J.,McGrath D. (1998) Oregon Cover Crops: Cereal Rye, Oregon StateUniversity Extension Service, EM 8694.

Schabenberger O., Pierce F.J. (2002) Contemporary statistical models forthe plant and soil sciences. Chapter 5. Nonlinear models, CRC Press,Boca Raton, Florida.

Schutter M.E., Dick R.P. (2002) Microbial community profiles and ac-tivities among aggregates of winter fallow and cover-cropped soil,Soil Sci. Soc. Am. J. 66, 142–153.

Teasdale J.R., Devine T.E., Mosjidis J.A., Bellinder R.R., Beste C.E.(2004) Growth and development of hairy vetch cultivars in thenortheastern United States as influenced by planting and harvest-ing date, Agron. J. 96, 1266–1271.

Thorup-Kristensen K. (2001) Are differences in root growth of nitrogencatch crops important for their ability to reduce soil nitrate-N con-tent, and how can this be measured? Plant Soil 230, 185–195.

Tsoularis A., Wallace J. (2002) Analysis of logistic growth models, Math.Biosci. 179, 21–55.

Vidal M., López A. (2005) Cover crops and organic amendments to pre-vent nitrate contamination under a wet climate, Agron. Sustain.Dev. 25, 455–463.

Werker A.R., Jaggard K.W. (1997) Modelling asymmetrical growthcurves that rise and then fall: applications to foliage dynamics ofsugar beet (Beta vulgaris L.), Ann. Bot. 79, 657–665.

Yin X., Goudriaan J., Lantinga E.A., Vos J., Spiertz H.J. (2003) A flexiblesigmoid function of determinate growth, Ann. Bot. 91, 361–371.

Zuazo V.H.D., Pleguezuelo C.R.R. (2008) Soil-erosion and runoff pre-vention by plant covers. A review, Agron. Sustain. Dev. 28, 65–86.

improved evaluation of cover crop species by growth and

Documents